Modern phones pack a good bit of compute, and can run things like VLAs decently well.
Of course, that would require today's phones to age out of "being used as a phone" bracket, and robotics VLAs to become actually useful. But things like the Comma AI autopilot hardware use slightly obsolete smartphone chips internally - so it's not like it's impossible to run a useful AI on this kind of HW.
Apple only uses Qualcomm chips as modems. Almost everyone else uses Qualcomm chips as main SoCs.
Now, could hardware vendors tell Qualcomm to go pound sand and run their own support for old SoCs? Yes they could. Do they want to? Hell no, supporting old devices doesn't make any money.
I think expecting software licenses to enforce your rights outside of the realm of software is a pretty bad take. I think Linus's take is quite solid: "I give you my source code, you give me your changes back, and we're even". There are a lot of us who don't think that FOSS should be weaponized as a poison pill to enact the authors worldview on topics outside of the realm of software alone.
If it should be a consumer right, why limit it only to devices certain types of software? Why not consumer protection law that applies to all devices? I think software licenses are the wrong tool for this problem.
There's a lot of crazy crayon licenses out there that try to fix the whole world by tacking on a whole lot of restrictions to their software licenses, prohibiting use for a long list of reasons... to me it sounds like a bunch of newspeak, as if "more restrictions = more freedom"
Is being able to replace software on the devices I own not in "the realm of software" somehow?
"Sure, you can have the sources, you just can't use them on your own devices because the vendor that shipped it has decided to bar you from doing that with a 2048-bit RSA key" just feels like GPL was upheld in letter, but not in spirit.
Yeah, it is -- you're asking for restrictions on pieces of hardware unrelated to the original software other than the fact that someone decided to install it on there.
How would you feel if a piece of hardware came with a license prohibiting software developers from using encryption to secure their systems?
The root of the issue here is that phone hardware landscape is effectively a duopoly. It is an antitrust issue. Trying to use software licenses to do this 1) won't be effective because the duopoly will never use them, and 2) is like going around your ass to get to your elbow. Even if it did work it wouldn't get to the root of the issue. The law needs to fix the fact that almost all phones on the planet are controlled either directly or indirectly by two companies.
There will always be more ways for companies to extract value without contributing. Linus would have to continuously upgrade licenses from GPLV2 to V3 to Affero and so on. It is not really practical.
What Linus has contributed is already huge. We can't put all the burden of making the world right on him.
"Good enough" open weights models were "almost there" since 2022.
I distrust the notion. The bar of "good enough" seems to be bolted to "like today's frontier models", and frontier model performance only ever goes up.
I don’t see why. Today frontier models are already 2 generations ahead of good enough. For many users they did not offer substantial improvement, sometimes things got even worse. What is going to happen within 1 year that will make users desire something beyond already working solution? LLMs are reaching maturity faster than smartphones, which now are good enough to stay on the same model for at least 5-6 years.
Any considerable bump in model capability craters my willingness to tolerate the ineptitude of less capable models. And I'm far from being alone in this.
Ever wondered why those stupid "they secretly nerfed the model!" myths persist? Why users report that "model got dumber", even if benchmarks stay consistent, even if you're on the inference side yourself and know with certainty that they are actually being served the same inference over the same exact weights on the same hardware quantized the same way?
Because user demands rise over time, always.
Users get a new flashy model, and it impresses them. It can do things the old model couldn't. Then they push it, and learn its limitations and quirks as they use it. And then it feels like it "got dumber" - because they got more aggressive about using it, got better at spotting all the ways it was always dumb in.
It's a treadmill, and you pretty much have to keep improving the models just to stay ahead of user expectations.
I have seen this with ChatGPT progression from 4o to 5.2 applied to the newest model. Old prompts stop working reliably, different hallucination modes etc.
Methane has good energy density, doesn't demand cryogenics or diffuse through steel, burns very cleanly, and can be used in modified gasoline ICEs - without even sacrificing the gasoline fuel capability.
Isn't the point that it is as simple and convenient as normal gasoline and also that you can use your gasoline car? If you are using gases it is a hassle for everyone and you need a new car or a full retrofit. At some point we have to ask ourself why we would even do that. Is it really worth it compared to just using a battery?
Without cryogenics, methane has such low energy density that a low-pressure fuel tank would still have to be as big as a bus for your compact methane-powered vehicle to go as far as you could on a few gallons of gasoline.
In general, "green hydrogen" makes the most sense if used as a chemical feedstock that replace natural gas in industrial processes - not to replace fossil fuels or be burned for heat.
On paper, hydrogen has good energy density, but taking advantage of that in truth is notoriously hard. And for things that demand energy dense fuels, there are many less finicky alternatives.
(Not GP) There was a well recognized reproducibility problem in the ML field before LLM-mania, and that's considering published papers with proper peer-reviews. The current state of afairs in some ways is even less rigourous than that, and then some people in the field feel free to overextend their conclusions into other fields like neurosciences.
We're in the "mad science" regime because the current speed of progress means adding rigor would sacrifice velocity. Preprints are the lifeblood of the field because preprints can be put out there earlier and start contributing earlier.
Anthropic, much as you hate them, has some of the best mechanistic interpretability researchers and AI wranglers across the entire industry. When they find things, they find things. Your "not scientifically rigorous" is just a flimsy excuse to dismiss the findings that make you deeply uncomfortable.
Strange that they raised money at all with an idea like this.
It's a bad idea that can't work well. Not while the field is advancing the way it is.
Manufacturing silicon is a long pipeline - and in the world of AI, one year of capability gap isn't something you can afford. You build a SOTA model into your chips, and by the time you get those chips, it's outperformed at its tasks by open weights models half their size.
Now, if AI advances somehow ground to a screeching halt, with model upgrades coming out every 4 years, not every 4 months? Maybe it'll be viable. As is, it's a waste of silicon.
The prototype is: silicon with a Llama 3.1 8B etched into it. Today's 4B models already outperform it.
Token rate in five digits is a major technical flex, but, does anyone really need to run a very dumb model at this speed?
The only things that come to mind that could reap a benefit are: asymmetric exotics like VLA action policies and voice stages for V2V models. Both of which are "small fast low latency model backed by a large smart model", and both depend on model to model comms, which this doesn't demonstrate.
In a way, it's an I/O accelerator rather than an inference engine. At best.
Even if this first generation is not useful, the learning and architecture decisions in this generation will be. You really can't think of any value to having a chip which can run LLMs at high speed and locally for 1/10 of the energy budget and (presumably) significantly lower cost than a GPU?
If you look at any development in computing, ASICs are the next step. It seems almost inevitable. Yes, it will always trail behind state of the art. But value will come quickly in a few generations.
maybe they're betting on improvement in models to plateau, and that having a fairly stablized capable model that is orders of magnitude faster than running on GPU's can be valuable in the future?
Of course, that would require today's phones to age out of "being used as a phone" bracket, and robotics VLAs to become actually useful. But things like the Comma AI autopilot hardware use slightly obsolete smartphone chips internally - so it's not like it's impossible to run a useful AI on this kind of HW.
reply